Detecting anomalous gene clusters and pathogenicity islands in diverse bacterial genomes

Detecting anomalous gene clusters and pathogenicity islands in diverse bacterial genomes

Review TRENDS in Microbiology Vol.9 No.7 July 2001 76 Anderson, J.M. et al. (1990) Ultrastructure and antigenicity of the unique cell and pimple of ...

524KB Sizes 1 Downloads 45 Views

Review

TRENDS in Microbiology Vol.9 No.7 July 2001

76 Anderson, J.M. et al. (1990) Ultrastructure and antigenicity of the unique cell and pimple of the Candida albicans opaque phenotype. J. Bacteriol. 172, 224–235 77 Morrow, B. et al. (1993) Coordinate regulation of two opaque phase specific-genes during white–opaque switching in Candida albicans. Infect. Immun. 61, 1823–1828 78 Srikantha, T. and Soll, D.R. (1993) A whitespecific gene in the white-opaque switching system of Candida albicans. Gene 131, 53–60 79 Srikantha, T. et al. (1995) Functional analysis of the promoter of the phase-specific WH11 gene of

Candida albicans. Mol. Cell. Biol. 15, 1797–1805 80 Rustchenko-Bulgac, E.P. et al. (1990) Chromosomal rearrangements associated with morphological mutants provide a means for genetic variation of Candida albicans. J. Bacteriol. 173, 7436–7442 81 Perez-Martin, J. et al. (1999) Phenotypic switching in Candida albicans is controlled by a SIR2 gene. EMBO J. 18, 2580–2592 82 Sonneborn, A.B. et al. (1999) Control of white–opaque phenotypic switching in Candida albicans by the Efg1p morphogenetic regulator. Infect. Immun. 67, 4655–4660

335

83 Soll, D. (1988) High-frequency switching in Candida albicans and its relationship to vaginal candidiasis. Am. J. Obstet. Gynecol. 158, 997–1001 84 Jones, S. et al. (1994) Increased phenotypic switching in strains of Candida albicans associated with invasive infections. J. Clin. Microbiol. 132, 2869–2870 85 Kvaal, C. et al. (1999) Missexpression of the opaque-phase specific gene PEP1 (SAP1) in the white phase of Candida albicans confers increased in a mouse model of cutaneous infection. Infect. Immun. 67, 6652–6662

Detecting anomalous gene clusters and pathogenicity islands in diverse bacterial genomes Samuel Karlin A gene in a genome is defined as putative alien (pA) if its codon usage difference from the average gene exceeds a high threshold and codon usage differences from ribosomal protein genes, chaperone genes and proteinsynthesis-processing factors are also high. pA gene clusters in bacterial genomes are relevant for detecting genomic islands (GIs), including pathogenicity islands (PAIs). Four other analyses appropriate to this task are +C genome variation (the standard method); genomic signature divergences G+ (dinucleotide bias); extremes of codon bias; and anomalies of amino acid usage. For example, the cagA domain of Helicobacter pylori is highly deviant in its genome signature and codon bias from the rest of the genome. Using these methods we can detect two potential PAIs in the Neisseria meningitidis genome, which contain hemagglutinin and/or hemolysin-related genes. +C variation and genome signature differences of the Additionally, G+ Mycobacterium tuberculosis genome indicate two pA gene clusters.

Samuel Karlin Dept of Mathematics, Stanford University, Stanford, CA 94305-2125, USA. e-mail: karlin@ math.stanford.edu

Pathogenicity islands (PAIs) contain genes involved in diseases, such as genes encoding invasins, adhesins and secretion factors, and are often sources of toxins. In many cases, genes found in PAIs are similar to the genes in other pathogens that encode subunits of type IV secretion systems, which deliver bacterial virulence factors to the surface of, and possibly into, host cells. PAIs are a subset of genomic islands (GIs) comprising linked blocks of genes, such as the cobalamin operon of Salmonella typhimurium1 and iron-uptake systems in Yersinia spp.2, which are required for fitness and survival. However, under adverse conditions they can contribute to pathogenicity. PAIs are associated with virulence functions that are present in some closely related species but not in others. GIs have generally been found to differ significantly in G+C frequency from the average genome G+C frequency. Here, we http://tim.trends.com

describe new methods (Box 1) for discerning GIs and PAIs: by the genome signature profile (Box 2); by assessing codon biases from the complete genome (Box 3); and by identifying extremes of amino acid usages in the proteome (Box 4). PAIs and GIs

Genes that deviate substantially in codon usage from the average gene but that are similar in codon usage to ribosomal protein genes (RPs) tend to be ‘highly expressed’. By these criteria, highly expressed genes in most prokaryotic genomes include RPs, translation and transcription processing factors (TF), and chaperone and degradation protein complexes (CH)3. By contrast, a gene is deemed putative alien (pA) if its codon usage difference relative to the average gene in the genome (Box 3) exceeds a high threshold and if its codon usage differences from RPs, chaperone genes and protein synthesis genes are also high. pA genes include: many open reading frames (ORFs) of unknown function (many of which are probably recent lateral transfers); genes acquired via conjugative plasmid integration (e.g. restriction/modification enzymes or neutralizers of antibiotics); cryptic prophages; transposases; insertion sequence (IS) elements; pilus and fimbrial genes; lipopolysaccharide (LPS) biosynthesis genes; and some genes expressed under particular environmental conditions. PAIs have been identified in, inter alia, some Escherichia coli strains, S. typhimurium, Vibrio cholerae, Yersinia sp., Pseudomonas aeruginosa and Helicobacter pylori. Recent reviews have discussed

0966-842X/01/$ – see front matter © 2001 Elsevier Science Ltd. All rights reserved. PII: S0966-842X(01)02079-0

336

Review

TRENDS in Microbiology Vol.9 No.7 July 2001

Box 1. Genomic characterizations of anomalous gene regions Take a sliding window W (length 10, 20...50 kb) and evaluate the following: • Compositional contrasts (standard method): compare G+C frequency within W to the average genomic G+C frequency. • Genome signature contrasts: compare δ* differences of each window segment to the average genomic signature. • Codon usage contrasts: compare codon biases of the gene set of each window to the average gene codon usages. • Amino acid contrasts: compare amino acid biases of proteins in each window relative to the average proteome amino acid frequencies. • Putative alien (pA) gene clusters: compare differences in codon usages from the RP, TP and CH gene classes, and from the average gene.

the potential mechanisms of bacterial pathogenicity and the acquisition of PAIs by lateral gene transfer4–7. These authors mainly describe primary examples of PAIs in enterobacteria located at or near tRNA genes. Actually, tRNA-like elements are common sites for the integration of foreign DNA for prophage lysogenization, which might convert to virulence5. In enterobacteria, most PAIs are found

adjacent to selC (selenocysteine-specific tRNAsel)8. PAIs of 10–200 kb in length are often flanked by small direct repeats or IS elements and generally encode an integrase that is inserted into the bacterial chromosome5. Virulence is often abetted by the absence of specific resident genes, for example the gene encoding the surface protease OmpT, which attenuates pathogenicity7. Various major virulence genes are often incorporated into bacterial cells via plasmids and bacteriophages5; several PAIs are located on plasmids in Shigella and Yersinia8. A variety of toxins are encoded by bacteriophages, including Shiga toxin in enterobacteria, cholera toxin in V. cholerae and neurotoxins in some strains of Clostridium9. Plasmids carry restriction systems, antibiotic resistance genes, heavy metal cofactors, nif (nitrogen fixation) genes and other contingency functions5–7. Additionally, microbial genomics has also helped elucidate microbial evolution with respect to similarities and differences among PAIs, disease mechanisms, cellular localization, antigenic variation, host–microorganism interaction,

Box 2. Genome signature comparisons Every genome has its characteristic signature calculated from genomic sequencesa,b. Explicitly, the genome signature profile consists of the array of dinucleotide relative abundance values:



* XY

=f

* XY

}

* * X Y

/f f

where f *X denotes the frequency of the mononucleotide X and f *XY the frequency of the dinucleotide XY, both computed from the sequence concatenated with its inverted complement sequence. These { ρ*XY } (actually ρ*XY−1, termed dinucleotide biases) essentially assess differences between the observed dinucleotide frequencies and those expected from random associations of the component mononucleotide frequencies. Biochemical experiments in the 1960s and 1970s measuring nearest-neighbor frequenciesc–e established that the { ρ*XY } values present a remarkably stable (ten-component) array for the DNA of an organism. Genomic DNA samples (≥50 kb) from different chromosomal locations of the same genome have approximately constant signatures, and closely related species have more similar genome signatures than distantly related species. From this perspective, the {ρ*XY } array constitutes a genomic signature that is diagnostic and can discriminate among sequences from different organismsa,f,g. The mechanisms that generate and maintain the signature are not understood. A reasonable explanation postulates that these are the result of differences in the replication and repair machinery of different species, which either preferentially generate or preferentially select specific dinucleotides in the DNA. These effects might operate through local DNA structures (base-step stacking energies and conformational tendencies), context-dependent mutation rates, methylation and/or other DNA modifications. A measure of genomic signature difference between two sequences f and g (from different organisms or from

http://tim.trends.com

different regions of the same genome) is the average absolute dinucleotide relative abundance difference calculated as:

δ * (f, g) =

1 ∑ | ρ*XY (f) − ρ *XY (g) | 16

(termed δ* difference), where the sum extends over all dinucleotides. We propose that evolutionarily successful DNA integration including cell fusions requires that the genome signatures of the two partners be moderately similar, meaning δ* values ≤0.90 or, at least, weakly similarh (δ* ≤1.20). Plasmids (specific or of broad host range) and hosts tend to have moderately similar genome signaturesb. We speculate that laterally transferred genes evolve rapidly toward the signature of their new host. Concomitantly, we expect generally that G+C content, codon bias and amino acid usage adopt the genome-wide tendencies of these measurements. References a Karlin, S. (1998) Global dinucleotide signatures and analysis of genomic heterogeneity. Curr. Opin. Microbiol. 1, 598–610 b Campbell, A. et al. (1999) Genome signature comparisons among prokaryote, plasmid and mitochondrial DNA. Proc. Natl. Acad. Sci. U. S. A. 96, 9184–9189 c Josse, J. et al. (1961) Enzymatic synthesis of deoxyribonucleic acid VIII. Frequencies of nearest neighbor base sequences in dexoyribonucleic acid. J. Biol. Chem. 263, 864–875 d Russell, G.J. et al. (1976) Doublet frequency analysis of fractionated vertebrate nuclear DNA. J. Mol. Biol. 108, 1–23 e Russell, G.J. and Subak-Sharpe, J.H. (1977) Similarity of general designs of protochordates and invertebrates. Nature 266, 533–536 f Karlin, S. et al. (1997) Compositional biases of bacterial genomes and evolutionary implications. J. Bacteriol. 179, 3899–3913 g Gentles, A. and Karlin, S. (2001) Genome-scale compositional comparisons in eukaryotes. Genome Res. 11, 540–546 h Karlin, S. et al. (1999) A chimeric prokaryotic ancestry of mitochondria and primitive eukaryotes. Proc. Natl. Acad. Sci. U. S. A. 96, 9190–9195

Review

TRENDS in Microbiology Vol.9 No.7 July 2001

337

Box 3. Codon usage contrasts between different gene classes Let G be a family of genes with average codon frequencies g(x,y,z) for the codon nucleotide triplets (x,y,z) normalized so that:

∑ g (x,y,z) = 1

(x,y,z) = a

where the sum extends over all codons translated to amino acid a. Let f (x,y,z) indicate the average codon frequencies for the gene family F (F can be a single gene g) again normalized to 1 in each amino acid codon family. The codon usage difference of the gene family F relative to the gene family G, termed the codon bias of F with respect to G, is calculated by the formula: B (F|G) =





 (x,y,z) = a



∑ pa(F)  ∑ |f (x,y,z) − g (x,y,z)|  a

where {pa(F)} are the average amino acid frequencies of the gene family F (Refs a,b). For the collection of all genes of length at least 100 amino acids, which we designate C, B(F|C) is referred to as the codon bias of the gene set F relative to the average gene of the genome.

Let RP denote the collection of all ribosomal protein genes of the genome, CH the major chaperone/degradation genes and TF the principal protein synthesis genesc. Formally, a gene, g, is considered putative alien (pA) if the biases B(g|RP), B(g|CH), B(g|TF) and B(g|C) all exceed M + 0.15, where M is the median codon bias of B(g|C) over all genes. The statistical significance of pA gene clusters or clusters of predicted highly expressed genes is identified using the scan algorithms set forth in Ref. d; see also Ref. e on the distribution of uptake signal sequences in the H. influenzae genome. References a Karlin, S. et al. (1998) Codon usages in different gene classes of the E. coli genome. Mol. Microbiol. 29, 1341–1355 b Karlin, S. et al. (1998) Comparative DNA analysis across diverse genomes. Annu. Rev. Genet. 32, 185–225 c Karlin, S. and Mrázek, J. (2000) Predicted highly expressed genes of diverse prokaryotic genomes. J. Bacteriol. 182, 5238–5250 d Karlin, S. and Macken, C. (1991) Some statistical problems in the assessment of inhomogeneities of DNA sequence data. J. Am. Stat. Assoc. 86, 27–35 e Karlin, S. et al. (1996) Frequent oligonucleotides and peptides of the Haemophilus influenzae genome. Nucleic Acids Res. 24, 4263–4272

characteristics of virulence factors and information pertinent to drug discovery. PAIs can differ from the rest of the genome in several ways. Five methods are proposed for detecting PAIs and GIs: (1) G+C content anomalies (the standard method). Take a sliding window of 10, 20...50 kb in length and ascertain the average G+C frequency of each, singling out windows significantly deviant in G+C content. (2) Genomic signature contrasts (dinucleotide bias) (Box 2). The genome signature of each individual window is compared with the average genome signature over all windows of the genome. A highly disparate genome signature might imply a potential PAI. (3) Extremes of codon bias (Box 3). For each window W, collect all genes of the window and calculate the codon bias of FW with respect to the average gene of the genome. An outlying gene with high codon bias could be a potential PAI. (4) Divergence in amino acid usages (Box 4). For window W and genes FW , a window with strongly altered amino acid frequencies from the average could represent a PAI (Box 4). (5) pA gene clusters. Means for discriminating anomalous gene regions in many (but not all) cases are synonymous to identifying clusters of pA genes. For example, pA gene clusters identified using these methods include the Mu prophage region (1559–1593 kb) of Haemophilus influenzae, antigenic LPS biosynthesis gene clusters on the surface of many prokaryotic genomes, and the toxin-coregulated pili (tcp) gene cluster of V. cholerae (Fig. 1). These five methods are largely independent. The genome signature (i.e. the vector of dinucleotide relative abundances, Box 2) factors out the embedded mononucleotide frequencies. Codon-usage differences http://tim.trends.com

and genome signature evaluations are effectively uncorrelated assessments. Codon bias and amino acid differences are essentially complementary measures of gene composition. The methods outlined in Box 1 can be extended to higher-order oligonucleotide contrasts; for example, tetranucleotide relative abundances factor out mono-, di- and trinucleotide abundances, and we can also deal with dicodon biases. Putative GIs and/or PAIs in diverse bacterial genomes

We have applied the procedures outlined in Box 1 to detect potential PAIs or pA gene clusters in several genomes. Not every pA gene cluster is a PAI and, conversely, not every PAI is a pA gene cluster. H. pylori

H. pylori is a Gram-negative spiral-shaped bacterium that colonizes the human stomach. Approximately Box 4. Extremes of amino acid usages For the genes FW , we compare the amino acid frequencies of FW with the average amino acid frequencies of the genome. Windows containing genes with significant deviations in overall amino acid usages might signify a PAI. For two classes of genes, F and G, the amino acid bias is calculated by the formula:

A (F|G) =

20

(120) ∑ |ai(F) − ai(G) | i =1

where {ai (F)} is the average amino acid frequency of ai in F.

Review

338

TRENDS in Microbiology Vol.9 No.7 July 2001

(a) Vibrio cholerae chromosome 1

(b)

G+C content (%) (bars below the line indicate clusters of alien genes) 50 45 40 35 200

A

B

C

60 55 50 45 40

D

δ*differences with respect to genomic average ρ* values

200

100 0

Neisseria meningitidis

A

C

B

δ*differences with respect to genomic average ρ* values

100 0

Codon bias with respect to the average gene

400 200 0

400 200 0

Codon bias with respect to the average gene D

Amino acid bias with respect to the average gene

Amino acid bias with respect to the average gene 400

400

200 0

G+C content (%) (bars below the line indicate clusters of alien genes)

200 0.5Mb

1.0Mb

1.5Mb

2.0Mb

0

2.5Mb

0.5Mb

1.0Mb

1.5Mb

2.0Mb

A: LPS biosynthesis genes (rfb operon); B: ORFs and a hemolysin secretion protein homolog; C: toxin-coregulated pilus biosynthesis (TCP) operon and ORFs; D: ~50 kb region contains 13 alien ORFs (unknown function)

A: ORFs and two hemagglutinin/hemolysin related proteins; B: five frpC operon genes (similar to cytotoxin, involved in pathogenesis), two transposases, ORFs; C: ORFs, two hemagglutinin/hemolysin-related proteins, hemolysin activator hecB, one transposase; D: major cluster of ribosomal protein genes

(c) Pseudomonas aeruginosa

(d)

G+C content (%) (bars below the line indicate clusters of alien genes)

40 35 30 25 20

70 65 60 55 200

Rickettsia prowazekii

δ*differences with respect to genomic average ρ* values

200

100

100

0

0

Codon bias with respect to the average gene C D F 400 B E A 200 0 1Mb 2Mb 3Mb

5Mb

δ*differences with respect to genomic average ρ* values

Codon bias with respect to the average gene

400 200 0

G 4Mb

G+C content (%) (R. prowazekii has no alien genes)

6Mb

0.5Mb

1.0Mb

A: six probable bacteriophage proteins; B: cryptic bacteriophage Pf1; C: colicin-like toxin and anti-toxin, two fimbrial biogenesis genes, pili assembly chaperone, ORFs; D: four putative glycosyl transferases (possibly LPS biosynthesis), galE, one type II secretion, ORFs; E: LPS biosynthesis genes, ORFs; F: LPS biosynthesis (wbp operon), with inserted hisH2, hisF2; G: major ribosomal protein cluster including rpoABC

(e) Helicobacter pylori strain J99

(f)

Escherichia coli strain O157

G+C content (%)

G+C content (%)

45 40 35 30 200

δ*differences with respect to genomic average ρ* values

200

B

A

100 0

55 50 45 40

0

Codon bias with respect to the average gene

400 200 0

0.5Mb

δ*differences with respect to genomic average ρ* values B

100

1.0Mb

1.5Mb

400 200 0

Codon bias with respect to the average gene D

C

A 1Mb

2Mb

3Mb

E

4Mb

F

5Mb

A: cag region; B: mostly ORFs, two virB homologs (virulence), two DNA topoisomerase I genes

A: putative RTX family exoprotein (5188 aa length); B: prophage CP933V; C: prophage CP933Y; D: putative type III secretion genes and ORFs; E: ribosomal proteins; F: esc region (type III secretion)

(g) Mycobacterium tuberculosis

(h) Bacillus subtilis

G+C content (%) (bars below the line indicate clusters of alien genes) 80 75 70 65 60 δ*differences with respect to genomic average ρ* values 200 100 0 400 200 0 600 400 200 0

A

50 45 40 35

δ*differences with respect to genomic average ρ* values B

D

E

Codon bias with respect to the average gene A 400 B 200 0 Amino acid bias with respect to the average gene

Amino acid bias with respect to the average gene

2Mb

C

100 0

Codon bias with respect to the average gene

1Mb

G+C content (%) (bars below the line indicate clusters of alien genes)

3Mb

4Mb

A: three PPE genes and two PE-PGRS genes; B: five PE-PGRS genes, ilvX (acetolactate synthase), fadD18 (acylCoA synthetase homolog)

400 200 0

1Mb

2Mb

3Mb

4Mb

A: major cluster of ribosomal protein genes surrounded by four rRNA/tRNA operons; B: mostly ORFs, four putative transcriptional regulators; C: mostly ORFs, three putative regulators (one labeled phage related); D: ~150 kb region of mostly ORFS (six are labeled phage related); E: ~100 kb region of mostly phage related ORFs, several transport proteins

TRENDS in Microbiology

http://tim.trends.com

Review

Fig. 1. The contrasts are displayed for 20 and 50 kb sliding window plots of G+C content, genomic signature profiles, codon usage biases and amino acid anomalies for eight different bacterial species. Abbreviations: LPS, lipopolysaccharide; ORF, open reading frame.

TRENDS in Microbiology Vol.9 No.7 July 2001

50% of all humans are infected with H. pylori but only 10% exhibit clinical diseases including gastric carcinoma and peptic ulcer10. Three regions stand out in the G+C plot of the H. pylori strain 26659 genome: A (from 445 to 475 kb), B (545–585 kb) and C (1048–1068 kb). Patients with peptic ulcer often express the genes in region B, referred to as the cytotoxin-associated gene (cagA) island. The genomes of two H. pylori strains, 26695 and J99, have been sequenced in their entirety11,12. Virulent H. pylori differ from less-virulent strains by the presence of the ~40-kb segment of genes corresponding to the cagA locus. The cagA region of H. pylori deviates the most in genomic signature from the rest of the genome; comparison of the codon bias of the genes in each 40 kb interval with respect to the average H. pylori gene shows that genes in the cagA region carry the highest codon bias. Strains including this region cause cultured gastric epithelial cells to secrete the proinflammatory cytokine interleukin (IL)-8; this ability is abolished by specific mutation of many of the 26 ORFs found in the PAI (Ref. 13). The two isolates of H. pylori strains 26695 and J99 (Fig. 1e) feature three and two outlying domains, respectively, with a G+C frequency that is approximately 35%, roughly 4% lower than in the rest of the genome. In the J99 genome, this ‘plasticity zone’ unites regions A and C of the 26695 genome and qualifies as a GI when analyzed by three methods (G+C content, δ* deviants and codon biases). This region was named the plasticity zone as it contains nearly 50% of the genes that are unique to strains 26695 and J99. The plasticity zone contains two virB homologs (virulence factors) and two DNA topoisomerase I genes14.

339

V. cholerae

In the large chromosome of V. cholerae15 a massive pA gene cluster of ~50 kb spans a PAI (C in Fig. 1a) that contains the tcp region. V. cholerae Tcp is a type IV pilus that is an essential intestinal epithelium adherence factor. Two other pA gene clusters occur at 197–220 kb and 265–295 kb. Four of the methods discussed in Box 1 identify a significant pA cluster (D in Fig. 1a), highlighting 13 alien ORFs. The genes encoding the cholera enterotoxins CtxA and CtxB, both around position 1500 kb in chromosome I, were not detected by our methods but are both pA genes. The receptor for entry of the Ctx prophage into the V. cholerae cell is contained within the tcp gene conglomerate. Genes encoding potential toxins include secreted hemolysins (region B), proteases and lipases. A gene cluster (covering 17 kb length) contains several LPS biosynthesis genes (region A). The V. cholerae large chromosome contains two significantly large segments in antipodal positions: H1 = 43 − 327 kb and H 2 = 1657 − 1985 kb

devoid of highly expressed genes, and contains intervening regions devoid of pA genes: A1 = 2196 −2573 kb, A2 = 2621 −2885 kb and A3 = 1168 −1437 kb

These indicate that highly expressed and alien genes are irregularly distributed in the V. cholerae chromosomes. Neisseria meningitidis

All methods confirm two potential PAI peaks (A and C) in the N. meningitidis genome, emphasizing pairings of hemagglutinin and hemolysin-related genes (Fig. 1b)16. P. aeruginosa strain PAO1

0.80

1.40

ρ*(TA)

0.60

1.20

0.40

1.00

1.40 1.20 1.00 1.20

1.20 ρ*(GC) ρ*(AT)

1.00

ρ*(AA/TT)

A

ρ*(CC/GG)

B 1.00

1.00 1.00

At 6.3 Mb, the genome of this Gram-negative bacterium is the largest completely sequenced to date17 and is viewed as an opportunistic human pathogen. Examination of its codon bias (Fig. 1c) shows two cryptic regions (A and B) of bacteriophage genes. Peak C corresponds to genes encoding a colicinlike toxin and anti-toxin. Three distinct pA gene clusters (D, E and F) encode clusters of LPS biosynthesis genes (see also Fig. 3).

ρ*(CG)

0.80

Mycobacterium tuberculosis

1.20

ρ*(AG/CT)

0.80

1.00

0.60 1.20

0.80 ρ*(AC/GT)

ρ*(GA/TC)

1.20 ρ*(CA/TG)

1.00 1.00 0.80

1Mb

2Mb

3Mb

4Mb

1Mb

2Mb

3Mb

4Mb

TRENDS in Microbiology

Fig. 2. The ten individual component signature contrasts in Mycobacterium tuberculosis over a 50-kb sliding window. The distinguished regions A and B identified earlier are detectable for several individual signature component plots, strengthening confidence in their pathogenicity island (PAI) character. Dotted lines in the plots indicate genomic average ρ* values. http://tim.trends.com

M. tuberculosis, the causative agent of tuberculosis, is an intracellular pathogen of mononuclear phagocytes especially adapted to humans18. Genome signature differences and amino acid biases for 50 kb windows of the M. tuberculosis genome (Fig. 1f) show two outstanding segments, A (3730–3770 kb) and B (3925–3950 kb). There are around 80 unusual genes labeled PE-PGRS in the M. tuberculosis genome19, which encode a preponderance of glycine–glycine doublets, are significantly low in charged (acidic and basic) residues, aliphatic and aromatic residues, and virtually devoid of cysteine residues, thereby

340

Review

TRENDS in Microbiology Vol.9 No.7 July 2001

Genomes Eco Vch Hpy 40 59 173 36 164 36

Pathogenicity or genomic islands

Pae Nme 108 105 100 129 221 161 32 176 42

Mtu 110 127 271 90 174 33

121 141 275 164 178 115

Eco Vch 70 78 100 58 64 75 91 57 183 145 169 135 109 99 94 133 139 139 171 119 111 142 139 154

Hpy Pae 149 178 93 78 144 128 168 77 64 124 125 103 159 190 237 152 210 73 65 59 182 204 141 164 213 217 275 127 103 128

143 163 171 141 267 281 194 129 192 54 66 77 77 145 185 77 60 139 39 64 112 138 65 61 128 * 17 82 112 146 62 54 133 * 126 141 87 92 166 71 97 139 146 * 72 149 190 206 67 116 * 92 *

*

*

Nme 74 71 83 75 112 120 152 155 109 114 175 172

Mtu 150 183 173 214 239 287 195 258 140 169 135 196

164 95 65 98 37 124 129 102 119 179

185 144 190 196 174 247 280 192 190 230 170 164

*

156 84 73 97 39 124 135 106 109 176 22 *

*

159 211 244 252 198 308 330 250 226 297 207 199 137 *

E. coli O157 V. cholerae H. pylori J99 P. aeruginosa N. meningitidis M. tuberculosis Eco: rtx exoprotein (A) Eco: CP-933V prophage (B) Eco: esc region (F) Vch: tcp/acf region (C) Vch: otn region Hpy: cag region (A) Hpy: plasticity zone (B) Pae: colicin-like region (C) Pae: wbp region (F) Pae: psc region (type III secretion) Nme: hemolysin region (A) Nme: hemolysin region (C) Mtu: PPE/PE-PGRS region (A) Mtu: PPE region (B)

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300 Closely similar

Moderately similar

Weakly similar

Distantly similar

Distant

Very distant

TRENDS in Microbiology

Fig. 3. Average δ* differences among complete genomes and genomic islands based on 30-kb sequence samples. Letters in parentheses next to pathogenicity island descriptions indicate corresponding peaks in Fig. 1. The Vibrio cholerae otn region is associated with some pathogenic strains but missing from the current genome. The psc region of Pseudomonas aeruginosa is not associated with a strong peak in Fig. 1c. Abbreviations: Eco, Escherichia coli O157; Hpy, Helicobacter pylori J99; Mtu, Mycobacterium tuberculosis.; Nme, Neisseria meningitidis; Pae, Pseudomonas aeruginosa; Vch, Vibrio cholerae.

repeats in toxin (RTX) genes, which are active after post-translational fatty acid acylation and transported to the cell surface21. Two prophage regions can be distinguished: CP-933V (B) and CP-933Y (C), and two peaks representing putative type III secretion genes. As expected, the K12 plots show no peaks (not shown).

precluding electrostatic, hydrophobic and disulfidebridge interactions. The two M. tuberculosis PAIs mainly contain PE-PGRS genes. There is evidence to suggest that the proteins encoded by these genes, which have anomalous repetitive structures, are surface exposed and can obstruct the host immune system. Only region B significantly deviates in G+C content. To our knowledge, neither A nor B has been investigated as a PAI (Fig. 2).

Statistics of pA gene clusters in several bacterial genomes

Obligate intracellular parasites

The Rickettsia prowazekii, Chlamydia trachomatis and Chlamydia pneumoniae genomes, when analyzed by any of the five methods outlined in Box 1, do not contain a PAI or even a pA gene cluster (Fig. 1d). This might be as expected as the host putatively protects these bacteria from lateral gene transfer. Also, by our methods, Ureaplasma urealyticum does not contain a pA gene cluster. E. coli strainO157:H7

This bacterium is responsible for bloody diarrhea in many countries. E. coli O157:H7 has about 1.34 Mb of DNA that is absent from the laboratory strain K12, which is essentially non-pathogenic, and K12 has about 0.53 Mb that is absent from O157:H7 (Ref. 20). It is thought that most differences are due to lateral transfer. The plots of Fig. 1f show a peak (A) of http://tim.trends.com

As shown in Table 1, the highest percentage of pA genes is detected in Deinococcus radiodurans (254; 9% of all its genes) and the highest overall count is in E. coli (272; 7%). Strong contrasts in the numbers of pA genes are found when comparing the mycoplasma species Mycoplasma genitalium (5; 1%) and Mycoplasma pneumoniae (50; 8%). Not surprisingly, R. prowazekii is totally devoid of pA genes, consistent with its protected status as an obligate intracellular (human) parasite. Similarly, C. trachomatis and C. pneumoniae possess only 1% pA genes, but no pA gene cluster. Significant proportions of pA genes are detected in the genomes of Synechocystis PCC6803 (7%) and Pyrococcus horikoshii (7%). Borrelia burgdorferi contains the second-smallest number of pA genes (3; 0.3%). There are no pA transposases or IS genes in small parasitic or archaeal genomes, with the exception of Archaeoglobus fulgidus. By contrast, M. tuberculosis has 10 pA transposases. Intriguingly, Bacillus subtilis appears to have no transposases. Genomes with LPS biosynthesis alien clusters include, to date, E. coli, V. cholerae, D. radiodurans, P. aeruginosa, Synechocystis spp., Aquifex aeolicus, Thermotoga maritima, Methanobacterium thermoautotrophicum, Thermoplasma acidophilum, P. horikoshii and Holobacterium spp.

Review

TRENDS in Microbiology Vol.9 No.7 July 2001

Table 1. Some statistics on putative alien genes ≥100 amino acidsa Genome All (%)

Alien genes Alien gene Known functionb ORFs Transposon+ISc clusters of ≥5 genes

E. coli

272 (7%)

94

178

27

H. influenzae

60 (4%)

34

26

5

1

V. cholerae

175 (5%)

56

119

2

10

B. subtilis

172 (5%)

59

113

0

8

D. radiodurans

254 (9%)

4

82

172

25

0

0

0

0

0

C. trachomatis

9 (1%)

3

6

0

0

14 (1%)

4

10

0

0

M. genitalium M. pneumoniae B. burgdorferi

5 (1%)

4

1

0

0

50 (8%)

32

18

0

0 0

3 (0.3%)

0

3

0

T. pallidum

26 (3%)

10

16

0

0

H. pylori

27 (2%)

5

22

0

0

M. tuberculosis

158 (4%)

59

99

10

4

Synechocystis

189 (7%)

78

A. aeolicus

74 (5%)

31

T. maritima

64 (4%)

M. jannaschii M. thermoautotrophicum

56

4

43

0

3

37

27

1

2

28 (2%)

9

19

0

0

77 (5%)

21

56

0

3

A. fulgidus

133 (6%)

19

114

4

4

P. horikoshii

135 (7%)

3

132

0

0

89 (5%)

17

72

0

0

P. abyssi

111

aAbbreviations:

IS, insertion sequence; ORF, open reading frame. bCounts of known genes include putative identification. cTransposon and IS counts are included in the counts of known alien genes.

pA gene clusters in E. coli K12

The pA genes of E. coli include at least four classes of IS and/or transposase families – IS2, IS5, IS30 and IS186 – accounting for a total of 27 genes. These genes feature in >10 clusters, with each cluster including at least four consecutive genes. A four-gene cluster, in one orientation, extending approximately 4 kb over positions 567–571 kb is encoded on the leading strand. This might represent a laterally transferred operon. A cluster of eight genes (of mixed orientations) includes genes similar to those encoding the e14 integrase, the bacteriophage P21 integrase and the phage P22 repressor protein c2. Apparently, these genes arose as prophage incorporations. A pA gene cluster of six genes extending over positions 1416–1427 kb corresponds to the Rac-defective prophage22. A contiguous 11 pA gene cluster (2100–2111 kb) features five genes (rfbX, C, A, D and B) functioning in LPS biosynthesis. A ten pA gene cluster (3791–3804 kb) has similarities to the surface antigen LPS core biosynthesis or glucosyltransferase proteins carrying the designations rfaL, K, Z, Y, J, I, B, S and P. Another pA cluster of eight genes (4311–4321 kb), encoded from the lagging strand, encodes the PhnP, M, L, K, J, F and E proteins involved in http://tim.trends.com

alkylphosphonate uptake. There are several pA genes encoding restriction-modification enzymes, namely merA, merC and hsdS. Other pA genes have been identified by similarity to transposons of the Tn2501 resolvase family.

10

R. prowazekii C. pneumoniae

341

B. subtilis pA genes

The pA genes of B. subtilis feature contiguous genes similar to DNA-methyltransferase (cytosine specific) genes, extending over positions 654–661 kb: a DNA restriction enzyme, a DNA-entry nuclease (nucA) and a pA two-component response regulator (citT). Several B. subtilis pA genes are similar to some encoding various ATP-binding cassette (ABC) transporters of glutamine, of dipeptides, unspecified transporters and the spore coat protein B (cotB). H. influenzae pA genes

In H. influenzae, the 60 pA genes detected from the 1529 genes23 (lengths ≥100 amino acids) include 26 ORFs of unknown function and 34 genes of assigned function24. Eleven genes found in the region spanning 1557–1593 kb contain deposits of a cryptic Mu-like prophage. This region is detectable by all five methods and by the absence of the uptake signal sequence AAGTGCGGT, which is abundant in the rest of the H. influenzae genome25. Putative alien genes in H. influenzae include HindIII, HincII and HgiDI, genes of the restrictionmodification systems that might have been acquired through plasmid integration. Bracketing these restrictase/methylase genes are homologs of cryptic phages incorporating fragments of phages λ and P22 (whose natural habitats are enterobacteria), Mu phage, the L5 phage of Mycobacterium smegmatis and phages resident in other bacteria that are capable of invading H. influenzae26. These phage fragments flank the Trp operon. pA genes are present in two significant clusters in the region of the defective prophage Mu. In addition to these clusters, there are five pairs of overlapping genes, including three restriction systems, the thiM and thiD genes involved in thiamin biosynthesis and the hypothetical ABC transporter HI0354, which is adjacent to the hypothetical protein HI0355. D. radiodurans pA genes

Among the pA genes of D. radiodurans, a cluster of enzymes contributing to LPS biosynthesis stands out. LPS biosynthesis proteins are replete with repeat structures that could help confound host immune responses. Six Par proteins (which help in cell partitioning) are encoded by pA genes and three pA pilin type IV genes (DR0548, DR1232 and DR1233) have been identified. In D. radiodurans, we can identify 41 individual transposases (nine families) of which 28 qualify as pA genes27. The total number of pA genes in D. radiodurans is 282 (9% of all genes), of which 190 are ORFs of unknown function. A single extensive pA gene cluster on chromosome I consisting

342

Review

TRENDS in Microbiology Vol.9 No.7 July 2001

almost exclusively of ORF genes extends 27 kb and covers positions 527 kb to 554 kb (DR0526–DR0548). This region behaves as a GI. Signature differences among PAIs and their hosts

The average differences between 30 kb sequence samples of several genomes and PAIs of the same host and different PAIs are given in Fig. 3. It is striking that the genome signature differences of potential PAI regions and their host DNA are generally moderately similar to each other (the same degree of similarity as between plasmids and their hosts28). This similarity suggests that relatively close genome signatures promote PAI compatibility or that during their residence, the PAIs have changed relatively rapidly to their hosts’ signature. This analysis also indicates that the PAI sequences of E. coli O157:H7 and the two PAIs of V. cholerae are moderately similar in their genome signature (Fig. 3). Actually, the E. coli O157:H7 sequence and the Tcp gene cluster are remarkably close, δ* = 39. Strikingly, the genome signature of the PAI potential regions of P. aeruginosa is moderately similar, δ* = 59–73, to its host and to E. coli or V. cholerae but clearly distant or very distant from the other proteobacterial genomes. Similarly, the PAI cag region of H. pylori is weakly similar, δ* = 103, to its host. The two PAIs of N. meningitidis are close to the V. cholerae Otn region. The possible PAIs of M. tuberculosis are distant or very distant from all PAIs. Concluding comments

Acknowledgements I gratefully acknowledge Drs J. Mrázek, A. Campbell and B.E. Blaisdell of Stanford and H. Shuman of Columbia University Microbiology for valuable discussions. Supported in part by NIH grants 5R01GM10452-35 and 5R01HG00335-12.

Microbial genomes are substantially homogeneous in G+C content and genome signature29 but the composition of GIs differs, albeit not strongly, from that of the overall genome. Five criteria for detecting GIs and PAIs are highlighted in Box 1. When a genomic segment deviates sufficiently from the average genome composition in G+C, and/or genome signature (δ* distances), and/or codon usage, and/or amino acid frequencies and/or contains a cluster of pA genes, we view this gene segment as a GI, putatively derived by lateral gene transfer. If the same segment is identified using all five criteria we are more confident in the PAI status or GI character of that segment. Current studies of molecular evolution underscore lateral gene transfer as a major evolutionary mechanism7,30–33. For some genomes, it is estimated that 10–20% of genes have been laterally transferred within the last 100 million years34. Established vehicles for lateral transfer include transposition, conjugative plasmids, phage hitchhiking, free DNA

References 1 Lawrence, J.G. and Roth, J.G. (1996) Selfish operons: horizontal transfer may drive the evolution of gene clusters. Genetics 143, 1843–1860 2 Rakine, A. et al. (1999) The high pathogenicity island of Yersiniae. In Pathogenicity Islands and Other Mobile http://tim.trends.com

and natural transformation. Ochman et al.7 emphasize the dynamic character of genomes in balancing gene acquisition with gene loss. Accordingly, lateral gene transfer provides a means of bacterial diversification7. One would expect that genes beneficial to the recipient organism for its defense or in furnishing nutrients and metabolic addenda have potential for successful interspecies gene flow. However, successful incorporations are rare. We have previously verified that plasmids almost always have a genome signature moderately similar to that of any potential host28. I propose the following hypothesis. Genome signature similarity is essential for compatibility and transfer between two genomes or between gene domains and a host genome. In a parallel manner, we expect GIs to be close or moderately similar in genome signature to the host genome signature. One of the requirements for the successful acquisition of a laterally transferred gene is its utility to the recipient organism. For example, the heat shock (HSP60) proteins (GroEL and thermosomes) facilitate folding of a great variety of proteins and act on substrates whose selection seems to be solely constrained by their size35. The nonspecificity of HSP60 proteins relative to their targets makes them potential gene acquisitions. Along these lines, α-proteobacteria possess multiple features that suggest they can be donors of HSP60 genes: (1) they possess unusual facility for HSP60 gene duplication and transposition, as indicated by a plethora of HSP60 sequence replicates among classical α-proteobacteria, in contrast to the absence of paralogs of HSP60 sequences in the γ, β, δ and ε proteobacterial clades36; (2) α-proteobacteria establish close spatial and functional relationships, often endosymbiotic, with plant eukaryotic organisms37. Lateral gene transfer provides a means of rapid response to strong selection pressures induced by virulence factors ranging from toxin production to immune evasion. Examples include Ti plasmids of Agrobacterium tumefaciens, nodulation plasmids of Rhizobium and virulence plasmids of Shigella and Yersinia; (3) copies of α-proteobacterial HSP60 genes have been found to reside on extra chromosomal elements, for example the megaplasmids pSyma or Rhizobium meliloti. Ochman et al.7 describe examples of gene clusters acquired by lateral transfer that are near a host promoter and are expressed in the recipient cells. Although most PAIs are probably laterally transferred and have a compatible genome signature, it is not necessary that every GI is the same.

Virulence Elements (Kaper, J.B. and Hacker, J., eds), pp. 77–90, ASM Press 3 Karlin, S. and Mrázek, J. (2000) Predicted highly expressed genes of diverse prokaryotic genomes. J. Bacteriol. 182, 5238–5250 4 Finley, B.B. and Falkow, S. (1997) Common themes in microbial pathogenicity II. Mol. Biol. Microbiol. Rev. 61, 136–169

5 Hacker, J. and Kaper, J. (2000) Pathogenicity islands and the evolution of microbes. Annu. Rev. Microbiol. 54, 641−679 6 Kaper, J.B. and Hacker, J., eds (1999) Pathogenicity Islands and Other Mobile Virulence Elements, ASM Press 7 Ochman, H. et al. (2000) Lateral gene transfer and the nature of bacterial innovation. Nature 405, 299−304

Review

8 Parsot, C. and Sansonetti, P.J. (1999) The virulence plasmid of Shigellae: archipelagoe of pathogenicity islands. In Pathogenicity Islands and Other Mobile Virulence Elements (Kaper, J.B. and Hacker, J., eds), pp. 151–165, ASM Press 9 Iriarte, M. and Cornelis, G.R. (1999) The 70 kb virulence plasmid of Yersiniae. In Pathogenicity Islands and Other Mobile Virulence Elements (Kaper, J.B. and Hacker, J., eds), pp. 91–126, ASM Press 10 Blaser, M.J. and Parsonnet, J. (1994) Parasitism by the ‘slow’ bacterium Helicobacter pylori leads to altered gastric homeostasis and neoplasia. J. Clin. Invest. 94, 4–8 11 Tomb, J.F. et al. (1997) The complete genome sequence of the gastric pathogen Helicobacter pylori. Nature 388, 539–547 12 Alm, R.A. et al. (1999) Genomic sequence comparison of two unrelated isolates of the human gastric pathogen Helicobacter pylori. Nature 397, 176–180 13 Censini, S. et al. (1996) A cag, a pathogenicity island of Helicobacter pylori, encodes type Ispecific and disease-associated virulence factors. Proc. Natl. Acad. Sci. U. S. A. 93, 14648–14653 14 Liu, G. et al. (1999) Sequence anomalies in the cag7 gene of the Helicobacter pylori pathogenicity island. Proc. Natl. Acad. Sci. U. S. A. 96, 7011–7016 15 Heidelberg, J.F. et al. (2000) DNA sequence of both chromosomes of the cholera pathogen Vibrio cholerae. Nature 406, 477–483 16 Tettelin, H. et al. (2000) Complete genome sequence of Neisseria meningitidis Serogroup B Strain MC58. Science 287, 1809–1815

TRENDS in Microbiology Vol.9 No.7 July 2001

17 Stover, C.K. et al. (2000) Complete genome sequence of Pseudomonas aeruginosa PAO1, an opportunistic pathogen. Nature 406, 959–964 18 Orme, I.M. (1998) The immunopathogenesis of tuberculosis: a new working hypothesis. Trends Microbiol. 6, 94–97 19 Cole, S.T. et al. (1998) Deciphering the biology of Mycobacterium tuberculosis from the complete genome sequence. Nature 393, 537–544 20 Perna, N.T. et al. (2001) Genome sequence of enterohaemorrhagic Escherichia coli O157:H7. Nature 409, 529–533 21 Lally, E.T. et al. (1999) The interaction between RTX toxins and target cells. Trends Microbiol. 7, 356–361 22 Snyder, L. and Champness, W. (1997) Molecular Genetics of Bacteria, ASM Press 23 Fleischmann, R.D. et al. (1995) Whole-genome random sequencing and assembly of Haemophilus influenzae Rd. Science 269, 496–512 24 Mrázek, J. and Karlin, S. (1999) Detecting alien genes in bacterial genomes. Ann. New York Acad. Sci. 870, 314–329 25 Karlin, S. et al. (1996) Frequent oligonucleotides and peptides of the Haemophilus influenzae genome. Nucleic Acids Res. 24, 4263–4272 26 Hendrix, R.W. et al. (1999) Evolutionary relationships among diverse bacteriophage and prophage: all the world’s a phage. Proc. Natl. Acad. Sci. U. S. A. 96, 2192–2197 27 Karlin, S. and Mrázek, J. (2001) Predicted highly expressed and putative alien genes of Deinococcus radiodurans and implications for resistance to ionizing radiation damage. Proc. Natl. Acad. Sci. U. S. A. 98, 5240–5245

343

28 Campbell, A. et al. (1999) Genome signature comparisons among prokaryote, plasmid and mitochondrial DNA. Proc. Natl. Acad. Sci. U. S. A. 96, 9184–9189 29 Karlin, S. (1998) Global dinucleotide signatures and analysis of genomic heterogeneity. Curr. Opin. Microbiol. 1, 598–610 30 Campbell, A.M. (2000) Lateral gene transfer in prokaryotes. Theor. Popul. Biol. 57, 71−77 31 Brocchieri, L. (2001) Phylogenetic inferences from molecular sequences: review and critique. Theor. Popul. Biol. 59, 27−40 32 de la Cruz, I. and Davies, I. (2000) Horizontal gene transfer and the origin of species: lessons from bacteria. Trends Microbiol. 8, 128−133 33 Doolittle, W.F. (1998) You are what you eat: a gene transfer ratchet could account for bacterial genes in eukaryotic nuclear genomes. Trends Genet. 14, 307−311 34 Lawence, J.G. and Ochman, H. (1998) Molecular archaeology of the Escherichia coli genome. Proc. Natl. Acad. Sci. U. S. A. 95, 9413−9417 35 Houry, W.A. et al. (1999) Identification of in vivo substrates of the chaperonin GroEL. Nature 402, 147−154 36 Gupta, R.S. (1998) Protein phylogenies and signature sequences: a reappraisal of evolutionary relationships among archaebacteria, eubacteria, and eukaryotes. Microbiol. Mol. Biol. Rev. 62, 1435−1491 37 Ogawa, J. and Long, R. (1995) The Rhizobium meliloti groELc locus is required for regulation of early nod genes by the transcription activator NodD. Genes Dev. 9, 714−729

TIM online – making the most of your personal subscription High-quality printouts (from PDF files) Links to other articles, other journals, and cited software and databases All you have to do is: Obtain your subscription key from the address label of your print subscription. Then go to http://www.trends.com and click on the Claim online access button, then select Trends in Microbiology. You will see a BioMedNet login screen. Please enter your BioMedNet username and password. If you are not a BioMedNet member already, please click on the Join Now button to register. You will then be asked to enter your subscription key. Once confirmed, you have full-text access to Trends in Microbiology If you obtain an error message please contact Customer Services at [email protected] stating your subscription key and BioMedNet username and password. Please note you only need to enter your subscription key once, BioMedNet will remember your subscription. Institutional access is available at a premium; if your institute is interested in subscribing to print and online, please ask them to contact [email protected]

http://tim.trends.com